SensiML Eketsa Tlhokomelo ea Predictive ho Litaelo tsa Smart Building Devices

Hlahisa mehopolo le sepheo sa lab - metsotso e 10
Ts'ebetso ea "nako ea 'nete" ea mohato ka mohato oa ho theha mohlala - metsotso e 60
- Flash SensiML e sebeletsanang le pokello ea data firmware ho Thunderboard Sense 2 (TBS2)
- Lokisa le ho hokela TBS2 ho SensiML Data Capture Lab
- Tšoara lintlha tsa 'slide demo' ka boto e se nang letho (basebelisi ba ke ke ba ba le lisebelisoa tsa Fan)
- Ngola data ebe u boloka le sample project (ha re na ho sebelisa nako e setseng ea thupelo)
- Invoke Analytics Studio (ka nako ena, basebelisi ba tla be ba sebetsa ho tsoa ho dataset ea demo ea fan ea TBS2 e bokelletsoeng pele)
- Sebetsa ka mehato ea ho aha mofuta oa ho lemoha boemo ba fan
- Etsa Pakete ea Tsebo
- Ka boikhethelo: Moetso oa Flash ho ea ho TBS2

Video ea demo ea Smart Building Applications - metsotso e 5
Lipotso le Likarabo - metsotso e 10

SensiML Selelekela

SensiML ke k'hamphani ea lisebelisoa tsa software ea B2B bakeng sa AI moeling oa IoT
- E thusa ba ntlafatsang ho etsa mefuta ea lisensara tsa ML tsa ultra-compact ntle le tsebo ea mahlale a data
- Mehlala e nyane joalo ka 10KB!
- Sehlopha sa mehleng sa Intel Curie/Quark MCU AI sa lisebelisoa, se tlohetsoe ho theha SensiML ka 2017.
Silicon Labs le Tharollo ea SensiML
- Ho tlisa ML e sebetsang hantle ho lelapa la EFR32/EFM32 MCU
- Prototyping e potlakileng ea sesebelisoa sa IoT e nang le Thunderboard Sense 2

SensiML e na le botsitso le tšehetso ea lefats'e ka bophara
- E fumanoe ka 2019 ke QuickLogic Corp; seta le ho sebetsa e le setsi se ikemetseng sa software (e thehiloeng Portland, KAPA)
- Basebetsi-'moho le mocha ba thehiloeng (Avnet, Future Electronics, Mouser, Shinko Shoji)
- Liofisi tsa thekiso / litšehetso UK, US, Japan, Taiwan, China

Menyetla Bakeng sa TinyML Meahong e Bohlale

Mathata ka Nts'etsopele ea Ts'ebeliso ea Sensor ea Smart IoT e teng

Cloud-Centric AI

Sephethephethe se Phahameng sa Marang-rang

Boemo bo phahameng
Motho ea sa Mamelleng Liphoso hanyenyane
Kotsi e sa Tsejoeng ea Tšireletso ea Data
Matšoenyeho a Lekunutu

Thuto e Tebileng

Litlhoko tse kholo tsa data tsa koetliso
Setšoantšo se seholo sa memori
Mosebetsi o phahameng oa ho sebetsa
Tšebeliso e phahameng ea matla
Bophelo ba betri bo fosahetseng

Lintlha tsa ho Qetela ka letsoho

Ho lieha le ho sebetsa ka thata
Boholo ba khoutu bo sa tsejoeng ka pele
Tsebo ea mahlale a data a haellang
Lilaebrari tsa likhoutu tse rarahaneng tsa AI/ML

Ha se scalable/ tlholisano

TinyML = IoT Edge ML + AutoML

IoT Edge ML: Liphello tse ikemetseng
- Ts'ebetso e nyane ea marang-rang le bophelo bo bolelele ba betri ea waelese
- Ha ho ts'ebetso ea leru kapa ho itšetleha ka marang-rang
- Karabelo ea nako ea sebele

AutoML: Ntlafatsa Ntle le Tsebo ea AI
- Auto-optimizer e khetha mohlala o motle ka ho fetisisa bakeng sa data e fanoeng
- Ho ithuta ka mochini oa khale (ML) ho nyoloha ka ho ithuta ka botebo
- SensiML TinyML e hlahisa mefuta e menyane joalo ka 10KB!
Ho ngola ka letsoho ha ho hlokahale
- Khoutu ea mohlala e iketselitsoeng ho tsoa ho li-dataset tsa koetliso tsa ML
- E boloka likhoeli tsa boiteko ba nts'etsopele, le tsebo ea mahlale a data
- Moqapi a ka fetola karolo efe kapa efe ea khoutu ea AutoML kamoo a ratang

Mokhoa oa ho Haha oa Mohlala

Tšoara Lintlha

Nako: Lihora ho isa ho Libeke* (Ho ipapisitsoe le ho rarahana ha pokello ea data)
Tsebo: Tsebo ea Domain (Joalokaha ho hlokahala ho bokella le ho ngola liketsahalo tse khahlisang)

Hlokomela: Re tla sebelisa lintlha tse ling tse bokelletsoeng pele ho potlakisa mohato ona bakeng sa thupelo

Haha Mohlala

Nako: Metsotso ho isa ho Lihora (Ho ipapisitse le boemo ba taolo ea mohlala)
Tsebo: Ha ho letho (Full AutoML)
- Maele a mantlha a ML (Tokiso e tsoetseng pele ea UI)
- Python Programming (Taolo e felletseng ea lipeipi)

Sesebelisoa sa Teko

Nako: Metsotso ho isa ho Libeke (Ho ipapisitse le litlhoko tsa kopanyo ea khoutu ea app)
Tsebo: Ha ho na (Binary firmware e nang le khoutu ea wrapper ea I/O)
Embedding Programming (Ho kopanngoa ha laeborari ea SensiML kapa mohloli oa C o nang le khoutu ea mosebelisi)

Lipakane tsa Workshop

Tsebisa SensiML's TinyML toolkit le mokhoa oa ho aha mohlala ho Silicon Labs Thunderboard Sense 2.

Boiphihlelo ka nts'etsopele ea algorithm ea ML e tsamaisoang ke data
Ithute ts'ebetso ho tsoa pokellong ea lintlha ka netefatso le tlhahlobo ea sesebelisoa bakeng sa ho aha mefuta ea IoT
Theha mokhoa o sebetsang oa ho lokisa oa HVAC ho tloha qetellong

Arabela lipotso tseo u ka bang le tsona mabapi le ts'ebetso ea ho etsa mohlala oa TinyML

Kopo e sebetsang ea HVAC Predictive Maintenance Application

Bakeng sa merero ea karolo ea rona ea matsoho, re tla be re aha sesebelisoa se bohlale sa ho beha leihlo

Fans e sebelisitsoeng hohle ho aha lits'ebetso tsa HVAC: Li-blower, pholileng e sebetsang ea lisebelisoa, lisebelisoa tsa moea, mochini oa moea.
Ho hloleha kapa ho senyeha ho ka baka tahlehelo ea ts'ebetso, ho eketseha ha tšebeliso ea matla, ho hloleha ha HVAC
Re tla haha sesebelisoa se bonolo sa ho beha leihlo se ka bonang maemo a mangata a tloaelehileng le a sa tloaelehang:
- Tlosa / butle
- Lithaba tse hlephileng
- Tšitiso ea balateli
- Phallo ea moea e sa fellang kapa e koetsoeng ka botlalo
- Ho kenngoa ha lehare
- Ho thothomela ho feteletseng

Ha re Qalang Ts'ebetso

"Real-time" workshop mohato ka mohato mokhoa oa ho bōpa mohlala - metsotso e 60

Flash SensiML e sebeletsanang le pokello ea data firmware ho Thunderboard Sense 2 (TBS2)
Lokisa le ho hokela TBS2 ho SensiML Data Capture Lab

Tšoara lintlha tsa 'slide demo' ka boto e se nang letho (basebelisi ba ke ke ba ba le lisebelisoa tsa Fan)
Ngola data ebe u boloka le sample project (ha re na ho sebelisa nako e setseng ea thupelo)
Invoke Analytics Studio (ka nako ena, basebelisi ba tla be ba sebetsa ho tsoa ho dataset ea demo ea fan ea TBS2 e bokelletsoeng pele)

Sebetsa ka mehato ea ho aha mofuta oa ho lemoha boemo ba fan
Etsa Pakete ea Tsebo
Mohlala oa Flash ho TBS2

Video ea Demo

Litokomane / Lisebelisoa

SensiML Eketsa Tlhokomelo ea Predictive ho Lisebelisoa tsa Mohaho tse Smart [pdf] Litaelo
Kenya Tlhokomelo ea Pele ho Lisebelisoa tsa Meaho tse Smart, Tlhokomelo ho Lisebelisoa tsa Meaho e Bohlale, Lisebelisoa tsa Meaho e Bohlale, Lisebelisoa tsa Kaho

Litšupiso

Bukana ea Mosebelisi

SensiML Eketsa Tlhokomelo ea Predictive ho Lisebelisoa tsa Mohaho tse Smart

Lenaneo